A METHOD FOR ENHANCING ACCURACY IN FACE RECOGNITION SYSTEMS USING HYBRID NEURAL NETWORKS

Abstract

This article investigates a hybrid approach based on convolutional neural networks and Vision Transformer architectures to improve the accuracy of face recognition systems. The primary objective of the study is to enhance the quality and robustness of identification under challenging conditions, such as partial face occlusion (e.g., masks, glasses), low illumination, or images captured from angled viewpoints. The proposed hybrid model extracts local features through CNN layers and models global dependencies using Transformer blocks. By integrating local and global features, the approach aims to improve face recognition accuracy and overall system robustness.